Until you get the worst of both worlds and have to align an enterprise architecture approval committee, get cyber sign-off, put the request through your service desk, have it land with some sort of enterprise task prioritisation team, and finally land with your internal cloud team to action the request your team isn’t empowered to do, only to then have a simple service provisioning turn into a massively overcomplicated set of extra cloud bits and pieces getting provisioned and retemplated and secured and configured over the course of months, as that overallocated team find 20 minutes here or there only to hit some new wall as the cloud settings look different to the last time they did it or some unforeseen setting requirement was overlooked in the original request, leading to a two week lag on email responses and ticket handover before it gets picked up again.
That is why companies such as Spotify don't focus on cost savings as much as people here seem to think they should. A focus on cost savings inevitably means bureaucracy which means time spent on that instead of on product features. That means not just less growth for the stock markets but also more attrition due to good engineers hating the process. Of course at some point it's unsustainable but right until that moment it makes perfect sense.
edit: And that moment depends not just on costs but the growth rate of the company so it's almost impossible to predict ahead of time (otherwise wall street investing would be a lot more boring).
